Spotify says it has excluded more than 50 million tracks categorised as low-quality AI-generated content or spam from recommendations in 2026, without removing them from the service itself. The measure is distinct from outright deletion: in September 2025, the company reported having removed more than 75 million spam tracks for serious or repeated policy violations. Spotify says it continues to delete content on those grounds.
Over the past six months, the company has revised its recommendation system to give greater prominence to human artists. Its analysis found that tracks listeners skipped, ignored or explicitly rejected consistently overlapped with material flagged as low-quality AI content or spam. Spotify says listener complaints about such recommendations have subsequently declined. Detection relies on a combination of listening behaviour, upload patterns, metadata and third-party AI-detection tools. However, the company acknowledges that these tools have limitations and that identifying AI-generated material will become harder as the technology improves.
Identity labels form another part of Spotify’s approach. More than 300,000 human artists are now verified, accounting for over 99 per cent of artists users search for. Thousands of virtual artists carry an AI Persona label, indicating that they represent an artificial identity rather than a real person. Their music is excluded from editorial playlists and, by default, recommendations, although following an artist can create an exception. Artists also submit tens of thousands of credits disclosing AI use each day; listeners can view these in the mobile app’s Now Playing screen.
